在SQL Server中获取相互匹配

Question

I want to match up Buyers and Sellers. I'm doing this to make it as simple as possible.

I have a table: tmpSales. In it, I have a TransID, BuyerID, SellerID, Item and Date.

Sometimes buyers are sellers too, and vice versa.

What I want is a record where a buyer sold to a seller and that same seller sold to that same buyer.

I've got this query, which works fine:

SELECT *
FROM [dbo].[tmpSales] T1
INNER JOIN [dbo].[tmpSales] T2
ON T1.[BuyerID] = T2.[SellerID]
AND T2.[BuyerID] = T1.[SellerID]

However, it returns 2 records for each match. Is there any way for this to return a single record, with both BuyerID and SellerID present?

Sample data would look like so:

TransID     BuyerID     SellerID      ItemID      Date
1           10012       10032          65         10/15/2014
2           11111       10012          120        12/15/2014
3           10032       10012          32         2/2/2015
4           11111       10032          30         2/10/2015
5           10012       11111          45         3/1/2015  

In this case, I can see that 10012 and 10032 both sold to and bought from each other, as did 10012 and 11111 . I just want something like:

ID1             ID2
10012           10032
10012           11111

The data will be ever growing, so it's got to be dynamic (ie I can't put anything like, "Where BuyerID = '10012'" into the code).

EDIT: Actually, what I want to do is make this a view or stored procedure and pass 2 IDs to it, and have it tell me whether or not there is a mutual match.

Answer 1

You can easily limit the data in your select to be just one way by selecting which one is the first, like this:

SELECT *
FROM [dbo].[tmpSales] T1
INNER JOIN [dbo].[tmpSales] T2
ON T1.[BuyerID] = T2.[SellerID]
AND T2.[BuyerID] = T1.[SellerID]
AND T2.[BuyerID] > T1.[BuyerID]

So this way the one that has bigger buyer ID is always as the second one -- and you can have similar logic inside a procedure too.

Answer 2

You can use the following query to retrieve all the mutual map

SELECT DISTINCT T.BuyerID AS ID1,T.SellerID AS ID2
FROM [dbo].[tmpSales] T
WHERE EXISTS (SELECT * 
              FROM [dbo].[tmpSales] T1
              WHERE T1.BuyerID = T.SellerId
                    AND T1.SellerID = T.BuyerID)

If you want to check only for a specified pair, you can also do like this

DECLARE @Id1 INT
DECLARE @Id2 INT
SELECT *
FROM [dbo].[tmpSales] T
WHERE T.BuyerID = @Id1
AND T.SellerID = @Id2
AND EXISTS (SELECT * 
            FROM [dbo].[tmpSales] T1
            WHERE T1.BuyerID = @Id2
                AND T1.SellerID = @Id1)

Answer 3

This is in MySql but the same principles apply.

This query gets all of the distinct users for transactions made between two individuals where they've both been the buyer and the seller in different transactions. Since this includes all transactions each participant will be listed twice (once as the buyer and once as the seller -- and would be listed even more times with out the "distinct" if there was more than one sale in either direction).

select
    distinct sale1.buyer_id party1, sale1.seller_id party2
from
    temp_sale sale1
    join temp_sale sale2 on sale1.buyer_id = sale2.seller_id and sale1.seller_id = sale2.buyer_id
;

This gives these rows:

+ ----------- + ----------- +
| party1      | party2      |
+ ----------- + ----------- +
| 10032       | 10012       |
| 10012       | 11111       |
| 10012       | 10032       |
| 11111       | 10012       |
+ ----------- + ----------- +
4 rows

Adding the where clause below will get you the distinct set of individuals that have both bought and sold from each other.

select
    distinct sale1.buyer_id party1, sale1.seller_id party2
from
    temp_sale sale1
    join temp_sale sale2 on sale1.buyer_id = sale2.seller_id and sale1.seller_id = sale2.buyer_id
where 
    sale1.buyer_id > sale1.seller_id
;

This gives these rows:

+ ----------- + ----------- +
| party1      | party2      |
+ ----------- + ----------- +
| 10032       | 10012       |
| 11111       | 10012       |
+ ----------- + ----------- +
2 rows

For completeness, here is the entire example:

use example;

drop table if exists temp_sale;

create table temp_sale (
    id int,
    buyer_id int,
    seller_id int,
    item_id int,
    date date
);

insert into temp_sale values (1, 10012, 10032, 65, '2014-10-15');
insert into temp_sale values (2, 11111, 10012, 120, '2014-12-15');
insert into temp_sale values (3, 10032, 10012, 32, '2015-02-02');
insert into temp_sale values (4, 11111, 10032, 30, '2015-02-10');
insert into temp_sale values (5, 10012, 11111, 45, '2015-03-01');

select
    distinct sale1.buyer_id party1, sale1.seller_id party2
from
    temp_sale sale1
    join temp_sale sale2 on sale1.buyer_id = sale2.seller_id and sale1.seller_id = sale2.buyer_id
;

select
    distinct sale1.buyer_id party1, sale1.seller_id party2
from
    temp_sale sale1
    join temp_sale sale2 on sale1.buyer_id = sale2.seller_id and sale1.seller_id = sale2.buyer_id
where 
    sale1.buyer_id > sale1.seller_id
;

Answer 4

Aw... this was such an interesting question that 3 people got here before me. In any case, I recommend logging this data to a table as the dataset grows rather than just running an ad-hoc query each time you want to see it, but for the initial data population, here's a gross little cursor I scraped together. Hope it helps!

if object_id(N'tempdb..#FindMe') is not null
    drop table #FindMe ;

if object_id(N'tempdb..#UniqueRelationships') is not null
    drop table #UniqueRelationships ;

create table #UniqueRelationships
    ( Id1 int,
    Id2 int );

declare @Table table 
    ([TransID] int not null 
    ,[BuyerID] int not null 
    ,[SellerID] int not null 
    ,[ItemID] int not null 
    ,[Date] date not null );

insert @table values 
    (1,10012,10032,65,'2014-10-15 00:00:00.0'),
    (2,11111,10012,120,'2014-12-15 00:00:00.0'),
    (3,10032,10012,32,'2015-02-02 00:00:00.0'),
    (4,11111,10032,30,'2015-02-10 00:00:00.0'),
    (5,10012,11111,45,'2015-03-01 00:00:00.0');

select 
    t1.BuyerID
    ,t1.SellerID
into #FindMe 
from @table t1
join @Table t2 on t1.BuyerID = t2.SellerID
  and t2.BuyerID = t1.SellerID ;

declare
    @IdA int,
    @IdB int;

declare FindMe cursor for
    select BuyerId, SellerId from #FindMe ;

open FindMe ;
fetch next from FindMe into @IdA, @IdB ;
while @@fetch_status = 0
begin
    if not exists( select 1 
                    from #UniqueRelationships 
                    where ( Id1 = @IdA
                      and Id2 = @IdB ) 
                    or ( Id2 = @IdA
                      and Id1 = @IdB ) )
    begin
        insert #UniqueRelationships 
            ( Id1, Id2 )
        select 
            @IdA
            ,@IdB
    end

    fetch next from FindMe into @IdA, @IdB ;
end ;
close FindMe ;
deallocate FindMe ;

select * from #UniqueRelationships ;

if object_id(N'tempdb..#FindMe') is not null
    drop table #FindMe ;

if object_id(N'tempdb..#UniqueRelationships') is not null
    drop table #UniqueRelationships ;